IPv6 over gif(4) broken in 6.2-RELEASE?

[Dimitry Andric]

Bruce A. Mah wrote:
> Just to confirm, you're dealing with a gif(4) interface with an
> explicitly-configured destination address and a 128-bit prefixlen, yes?

Yes.  The specific line in my rc.conf is:

ipv6_ifconfig_gif0="2001:7b8:2ff:146::2 2001:7b8:2ff:146::1 prefixlen 128"


>> Maybe
>> there is something else involved too, for example the route command
>> itself?
> Not sure what you mean by this exactly...???...

I mean that it may be that between -RELEASE and -STABLE, other things
have changed, e.g. network rc scripts, /sbin/route itself, etc, which
may also influence this behaviour.  I'm sure more than only nd6.c
changed. :)

However, for me, with the whole system at -STABLE (as of Jan 11), I
verified the following results again just now:

nd6.c rev	state
---------	-----
1.48.2.12	works
1.48.2.13	works
1.48.2.14	works
1.48.2.15	works
1.48.2.16	doesn't work


> Here's what I've tested so far...in the table below, "work" means that
> the host route to the destination got installed correctly and "no work"
> means that it didn't.
> 
> Base		Local Patch			Result
> ----		-----------			------
> 6.2-RELEASE	Unmodified			No work
> 6.2-RELEASE	Revert nd6.c 1.48.2.{14,15}	Work
> 6.2-STABLE	Unmodified			No work
> 6.2-STABLE	Revert nd6.c 1.48.2.{14,15}	Work
> 6.2-STABLE	Revert nd6.c 1.48.2.16		No work

So strangely, this is different from my results... I can't install
6.2-RELEASE on the specific box, alas.


> I'm going to write up an entry for the 6.2-RELEASE errata notes
> documenting the existence of a problem and a workaround.  We still need
> to figure out exactly what the right fix is.  Testing results from other
> users (both 6.2-RELEASE and 6.2-STABLE) would be most welcome.

Just FYI, my initial alternative workaround was to use prefixlen 126,
e.g.:

ipv6_ifconfig_gif0="2001:7b8:2ff:146::2 prefixlen 126"

Cheers,
Dimitry